Aerospace and Electronic Systems - August 2018 - 61

Addabbo et al.

C
C

cdtu and cdts are respectively the receiver and satellite clock
offsets scaled by the speed of light,
Δdo is the orbital error,
Δdi and Δdt are respectively the ionospheric and tropospheric
delays,

Δdmp is the multipath error,

Δdn is the error related to receiver noise,

ε includes the unmodeled errors.

In SPP, the atmospheric errors are reduced using suitable models and the satellite clock offset is corrected using information
from the GNSS ground segment, included in the navigation message. The other errors remain uncorrected and would degrade the
SPP performance.
Equation (1) is then rearranged, comprising four unknowns,
i.e. the three user coordinates (included in the distance) and the
receiver clock offset, so at least four pseudo range equations are
necessary to estimate the unknowns. Equation (1) is not linear in
the unknowns, hence a linearization around a nominal state (consisting of approximate user position and clock offset) is usually
performed.
It is worth to notice that the SPP position has an accuracy of
about 5-10 meters [6] and so it is unusable in applications with
stringent requirements as mapping or precision farming.
3) Relative positioning: A typical approach to overcome SPP
shortage, regarding accuracy, is the relative or differential mode,
which makes use of one or more reference stations, placed in
known surveyed positions around the GNSS receiver with unknown coordinates (at maximum 100 Kilometers). Differential
mode, simply called Differential GNSS (DGNSS), is based on
the spatial correlation of several measurement error sources, i.e.
errors in the reference station measurements are expected to be
very similar to those experienced by a nearby user. The pseudorange error sources, cdts, Δdorb, Δdiono, Δdtropo, are spatially correlated and in the differential mode they are cancelled or strongly
reduced. In open-sky, where drones typically operate, the ionospheric error is the most influencing, so in this scenario the
DGNSS is very effective, improving the accuracy to 1-2 meters
[6]. To obtain better performance, such as centimeter-level positioning, the carrier-phase measurement should be used along with
pseudorange. The most common technique using carrier-phase
is Real Time Kinematic (RTK), which exploits the differential
concept too.
PR-based DGNSS technique could be implemented in either
position or measurement domain. In the first case, at the reference
station, it is computed the difference between the estimated and
the known position, which is broadcast to the user and applied directly as position correction. This approach is very simple, but it
does not consider the possible inconsistencies between the algorithms used by reference and user receivers. For this reason, it is
currently adopted when there is no access to user measurements,
such as in a cell phone using GPS. In measurement domain, at the
reference station, the differences between the measured and the
known ranges are computed and are broadcast to the user, which
AUGUST 2018

uses them as corrections for pseudoranges. This approach is very
common, and it is adopted for several applications as maritime
navigation.
The application considered herein, i.e. identification of defective photovoltaic panels by drones equipped with a thermal camera, requires a very accurate kinematic positioning, which currently can be obtained, in GNSS context, only using RTK technique.
In this work, the receiver based on RTK has not been available
from the beginning for problems of electromagnetic incompatibility with the antenna as built, so a PR-based DGNSS approach has
been followed. Moreover, the adopted GNSS device is unable to
provide raw data (e.g. pseduorange, carrier-phase, Doppler measurements), and so a position-domain DGNSS, based on National
Marine Electronics Association (NMEA) information [7], is implemented. In the proposed algorithm, the only data necessary from
the rover are GGA and GSA NMEA messages, while raw data are
required from the reference station. GGA NMEA message contains
the SPP coordinates of the rover, while GSA message contains the
identifiers of the satellites used for the solution.
The measurement corrections, computed at the reference station, are projected on the position domain using the design matrix.
In the considered test, a reference station placed at S. Nicola la
Strada (Caserta), far few Kilometers from the rover, is used. The
reference station belongs to the GNSS Campania Network and can
receive both GPS and GLONASS measurements.

THE IMAGE ACQUISITION SYSTEM: PV MODULE
IDENTIFICATION AND DEFECT DETECTION
The Image Acquisition procedure is based on a very precise processing system that receives images from the thermal camera in
input and produces final outputs through the proposed Computer
Vision Algorithm. As already highlighted geo-referencing operations are crucial since precision is the precondition of a correct
and satisfactory data processing for the entire system. Thus, it is
necessary, at the moment, the utilization of the DGPS technique
which improves accuracy of the present receiver, mounting on
board the U-blox NEO-M8N. As already mentioned, although
the U-blox NEO-M8N measures are less accurate with respect to
those of RTK, they introduce a very interesting novelty since initial
services of the Galileo constellation, supported by the NEO-M8N
GNSS module, have become available only recently. Moreover,
as already explained, DGNSS can compensate many of the error
sources by highly improving accuracy.
The core of the Image Acquisition System is the Computer
Vision Algorithm. It performs both PV module identification and
defect detection and gives a unique ID for each panel and a precise
relative position of defects in the metadata. Notably, we are interested in finding PV defectiveness and PV failure, defined as an
effect that results in safety or power loss for a PV module [8], [9].
The causes of defects in the PV panels can be damages occurred
during the planning, the transportation and the installation stages,
which could produce, for example, glass breakage, with a consequent loss of PV performance, or effects of adverse environmental
impacts. Among the main measurements methods used to identify

IEEE A&E SYSTEMS MAGAZINE

Aerospace and Electronic Systems - August 2018

Table of Contents for the Digital Edition of Aerospace and Electronic Systems - August 2018